Add better error handling and mock prediction fallback

app.py

@@ -20,31 +20,38 @@ def load_models():
         # Load models from your repository
         repo_id = "Nugget-cloud/nasa-space-apps-exoplanet"
         
+        # Try to get token from environment or use None for public repos
+        import os
+        token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACE_HUB_TOKEN")
+        
+        print(f"Loading models from {repo_id}...")
+        print(f"Using token: {'Yes' if token else 'No (public repo)'}")
+        
         print("Loading ensemble model...")
-        ensemble_model = joblib.load(hf_hub_download(repo_id, "exoplanet_ensemble_model.joblib"))
+        ensemble_model = joblib.load(hf_hub_download(repo_id, "exoplanet_ensemble_model.joblib", token=token))
         
         print("Loading feature scaler...")
-        feature_scaler = joblib.load(hf_hub_download(repo_id, "feature_scaler.joblib"))
+        feature_scaler = joblib.load(hf_hub_download(repo_id, "feature_scaler.joblib", token=token))
         
         print("Loading feature imputer...")
-        feature_imputer = joblib.load(hf_hub_download(repo_id, "feature_imputer.joblib"))
+        feature_imputer = joblib.load(hf_hub_download(repo_id, "feature_imputer.joblib", token=token))
         
         print("Loading variance selector...")
-        variance_selector = joblib.load(hf_hub_download(repo_id, "variance_selector.joblib"))
+        variance_selector = joblib.load(hf_hub_download(repo_id, "variance_selector.joblib", token=token))
         
         # Optional files
         try:
             print("Loading feature info...")
-            feature_info = joblib.load(hf_hub_download(repo_id, "feature_info.joblib"))
-        except:
-            print("Feature info not found, skipping...")
+            feature_info = joblib.load(hf_hub_download(repo_id, "feature_info.joblib", token=token))
+        except Exception as e:
+            print(f"Feature info not found: {e}")
             feature_info = None
             
         try:
             print("Loading model metrics...")
-            model_metrics = joblib.load(hf_hub_download(repo_id, "model_metrics.joblib"))
-        except:
-            print("Model metrics not found, skipping...")
+            model_metrics = joblib.load(hf_hub_download(repo_id, "model_metrics.joblib", token=token))
+        except Exception as e:
+            print(f"Model metrics not found: {e}")
             model_metrics = None
         
         print("All models loaded successfully!")
@@ -52,28 +59,52 @@ def load_models():
         
     except Exception as e:
         print(f"Error loading models: {str(e)}")
+        print(f"Repository: {repo_id}")
+        print("Make sure the repository exists and is public, or add HF_TOKEN to environment")
         return False
 
+def mock_predict(features):
+    """Fallback mock prediction when models can't be loaded"""
+    try:
+        # Simple mock logic based on feature values
+        feature_sum = sum(features)
+        prediction = 1 if feature_sum > 20 else 0
+        confidence = min(0.95, max(0.55, abs(feature_sum - 20) / 50 + 0.5))
+        
+        return {
+            "success": True,
+            "prediction": prediction,
+            "probabilities": [1-confidence, confidence] if prediction == 1 else [confidence, 1-confidence],
+            "confidence": confidence,
+            "input_features_count": len(features),
+            "note": "Using mock prediction - models could not be loaded",
+            "mock": True
+        }
+    except Exception as e:
+        return {
+            "success": False,
+            "error": f"Mock prediction failed: {str(e)}"
+        }
+
 def predict_exoplanet(features_input):
     """Make prediction using the loaded models"""
     global ensemble_model, feature_scaler, feature_imputer, variance_selector
     
     try:
-        # Load models if not already loaded
-        if ensemble_model is None:
-            if not load_models():
-                return {"error": "Failed to load models"}
-        
-        # Parse input features
+        # Parse input features first
         if isinstance(features_input, str):
-            # If input is comma-separated string
             features = [float(x.strip()) for x in features_input.split(',')]
         elif isinstance(features_input, list):
-            # If input is already a list
             features = [float(x) for x in features_input]
         else:
             return {"error": "Invalid input format. Expected comma-separated string or list of numbers."}
         
+        # Load models if not already loaded
+        if ensemble_model is None:
+            if not load_models():
+                print("Models failed to load, using mock prediction")
+                return mock_predict(features)
+        
         # Convert to numpy array
         features_array = np.array(features).reshape(1, -1)
         
@@ -112,10 +143,16 @@ def predict_exoplanet(features_input):
         return result
         
     except Exception as e:
-        return {
-            "success": False,
-            "error": str(e)
-        }
+        print(f"Prediction error: {str(e)}")
+        # Fallback to mock prediction
+        try:
+            features = [float(x.strip()) for x in features_input.split(',')] if isinstance(features_input, str) else features_input
+            return mock_predict(features)
+        except:
+            return {
+                "success": False,
+                "error": str(e)
+            }
 
 # Create a simple interface that works well with API calls
 def simple_predict(features_str):